Tulip AI does not do a good job in translation. In the 1st row, input field on the left contains info in both English and Spanish but the AI does nothing on that value and pastes same value as an output on the field on the right.
In the 2nd row, when the whole text was passed as Spanish, it then translates to English (however the keyword Pieza is still left untranslated).
Additional Info:
AI translation performed via “Automation” on Tulip table
Translated language: “EN-US”
Below are screenshots of Google translation.
Hey @akshay96 -
Thanks for reporting this. Under the hood we are using the DeepL service for translation and it is doing the source language detection. It looks like their support is limited for multi-language input. This model does not use a large language model under the hood (which has upside for things like GMP validation), but has limitation for problems like this. We did recently add
Google translate is using LLM’s to do translation, allowing them to better handles these mixed language cases.
I can achieve similar behavior by using the new “Prompt AI” action in automations. This reliably handles cases where the input is a mix of languages.
Hope this helps-
Pete
1 Like
Hey Pete, thanks a ton for your support. Although I am still seeing some minor issues but it is in a lot better state now then previously.
Thanks again 
